Multi-scale Arctic prediction with MPAS-CESM Nick Szapiro & Steven Cavallo AaARG, University of Oklahoma Bill Skamarock MMM, NCAR CESM Workshop 2017 PCWG Boulder, CO 2017-06-21 Grateful for support from ONR, NCAR MPAS-A group, & CISL 1/22x Arctic mesh
Mass, momentum, & energy deeply coupled across boundary layer, latitudes, & strat-trop MOSAiC Science Plan
Tropopause Polar Vortices Consider potential temperature on 2 PVU tropopause [e.g., Morgan & Nielsen-Gammon, 1998; Ivanova, 2013] TPVs are spatially and temporally coherent vortices O(100s-1000 km, 1-40 K, days-months) [Hakim & Canavan, 2005; Cavallo & Hakim, 2009] T v TPVTrack segmentation EPV RH
Upper level potential vorticity anomalies in synoptic IPV thinking Hoskins et al. 1985
With a spectrum of scales w/in each component: How does model design impact processes? How do filters impact evolution, variability, & prediction? MOSAiC Science Plan
Setup of CESM Large Ensemble [Kay et al. 2015]
Forced+internal variability in CESM-LE Greater sensitivity of thinner, younger ice [Maslanik et al. 2007; Kwok et al. 2013] With rapid ice loss events from anthro. forcing+growing intrinsic variability [Holland et al. 2008] March-September Arctic sea ice area from 30 members of CESM-LE [Kay et al. 2015] RCP8.5 Historical forcing
Sensible internal variability in CESM-LE: aice in 2021-09
(Area 2 ) (Area 2 ) Scales in CESM-LE: Power spectra of daily sea ice extent FFT-based power spectrum: 1. Daily SIE from aice.ge. 15% 2. Blackman taper 3. Mean PSD across members 1-30 PSD smoothed w/ +/- 5 sample mean NSIDC sea ice index 1989-2016 CESM-LE extent 1989-2016 Mem 1-30 w/ 95% Bootstrap C.I.
What drives rapid ice loss? Compositing procedure for synoptic events 18d High-pass pan-arctic SIE NSIDC concentration ERA-Interim Initially REU project w/ Uriel Gutierrez
Top 1% Synoptic Rapid Ice Loss Events: Associated w/ surface lows + TPVs Counts of loss objects for 25 summer cases Mean xice change 10 m wind Mean DT PT (color) and MSLP (contour) Composite mean in local reference frame w/ 10 m wind flowing to the right Composite mean in local reference frame centered on TPV core w/ MSLP min to the right
August 2006 SRILE Color: DT PT (K) Contour: MSLP.lt. 1000 hpa ( =4 hpa) AMSR-E
TPV-surface cyclone in WRF-ARW: Sensitivity to grid scale 120 km 90 km Color: RH (%) 24 km Contour: PT (K) 3 km Simulations: Chris Riedel
Polar filter affects TPV dynamics Increased damping of (Fourier) zonal wavenumbers towards pole for increased stability
Overview of Model for Prediction Across Scales (MPAS-A) MPAS-A Tutorial (http://www2.mmm.ucar.edu/projects/mpas/tutorial/uk2015/slides/mpas-overview.pdf)
MPAS-CESM 1.4.b7 (v2.0.b5 in testing) Dynamical, global, coupled, var-res (atmo) Atmosphere: CAM5.3 with MPAS-A dycore Physics: 30min dt w/ dribbled tendencies Vertical levels as in CAM5: 30 levels to 44.6km Land: CLM4 on ~1 FV grid 30 min coupling to atm Ocean: POP2 on ~1 Greenland tri-pole 1 day coupling to atm Sea ice: CICE4 on ~1 Greenland tri-pole 30 min coupling to atm River: RTM 3 hr coupling Coupling: CPL7 w/ atm * fluxes on atm grid Cost w/o optimization: ~1000 compute hrs/sim month ICs: analysis (atm) and CESM-LE analog (other components) MPAS-A 2 CICE 4 N/A CLM 4 RTM N/A POP 2 1/22x Arctic mesh
Refinement adds resolution in MPAS-CESM: f120h 500hPa rel. vertical vorticity 90 km 90-25 km Arctic 60 km 60 km 25 km 60 km
Use sensitivity experiments to quantify impacts of TPV intensity biases Localized tendency-based TPV modifications applied throughout integration ID region; e.g., DT PT<300K North of 70N Apply modification; e.g., dθ/dt -= 10K/day [DT-10K,DT)
Central Arctic TPVs can impact surface cyclones and sea ice transport: f72h DT PT (K) Control Strengthened TPV Ice velocity (cm/s)
Stronger TPVs at mid-latitudes can amplify waviness: f28d 2006 Control Intensify TPVs S of 65N
Stronger TPVs at mid-latitudes can impact sea ice motion: time-mean over 28d 2006 Control Intensify TPVs S of 65N
Options for ICs Cold start from external analysis Strongly coupled DA State accuracy Weakly coupled DA CESM-LE analog Forced by analysis Climatological Analytical Physical consistency
Initial conditions for 2017 Sea Ice Outlook SIO: Collection of June, July, and August 1 forecasts for September mean SIE for Arctic (, Alaska, and Antarctic) State accuracy Atm: Cold start from GFS FNL Other components: CESM-LE analog from 2021 2017* SIE Physical consistency Color: Sep. SIE in CESM-LE by member *linear extrapolation of Sep. Sea Ice Index
2017 Sea Ice Outlook and summary June SIO (M km^2): Arctic: 4.1 Alaska: 0.3 Antarctic: 18.1 TPVs can impact surface lows, wave amplification, seasonal circulation, and summer sea ice loss Future: Linearity and dynamics of coupled response to TPV intensity MPAS-CESM as a dynamical tool for process and prediction studies Future: Ensemble exploring model design and uncertainty Physics, mesh, numerics, coupling, stochastic physics, IC uncertainty: GEFS and CESM-LE analogs,?
Extra slides
Outline Background on polar prediction Coupled across BL, strat-trop, latitude Scale filters interactions CESM-LE Forced+natural variability in extent Coarse, polar filter, sea ice spectra MPAS-CESM description What is MPAS-A? Setup for MPAS-CESM Intrinsic predictability Impacts of weak synoptic features TPV modifications Practical predictability Forecast experiments 2017 SIO
Are CESM-LE integrations reproducible? aice for member 005 in 2021-09 Restart from 2021-01 Original Restart run on Yellowstone in May 2017 Bit-equal dependent on compilers, libraries, architecture,?
TPV dynamics: Ertel s Potential Vorticity Transport > Diabatic > Frictional Cavallo & Hakim 2008,2010,2013
Intrinsic and practical TPV predictability in IFS EPS Area of Θ <10 K Min(Θ)
Motivation for MPAS-CESM: Variable-res and coupled permits realism & flexibility Representing Arctic/TPVs is a coupled, multi-scale problem e.g., air-ice fluxes, Arctic polar jet, Global MPAS-A Numerics Transport, time integration, filtering, design guided by theory and practical experience in weather modelling [Skamarock et al. 2012] No polar filter Smooth variable horizontal refinement High-resolution can permit better representation of local processes TPV radii ~500km with finer sub-structures Limited area models suffer from boundaries (1) driving flow and (2) inconsistent with interior, worse for small and large domains, respectively [Laprise et al. 2008] Static refinement efficient for small Arctic CESM Representation and consistent evolution of atm, lnd, ice, ocn, 1/22x Arctic mesh